Diagnosing axle box bearings’ fault using a refined phase difference correction method

The wheelset treads and axle box bearings of railway vehicles often suffer from fatigue failures. Their regular maintenance highly depends on manual off-line inspection with low working efficiency and poor precision for early failure detection. This study proposes a fault diagnosis method by band-pass filtering and by enveloping the accelerations collected from the axle box bearing on the underfloor wheelset lathe to improve the maintenance efficiency. This process is followed by the refined phase difference correction using the four-term third derivative Nuttall-windowed fast Fourier transform (RPNWF) to extract accurate amplitudes of the fault characteristic frequency and its harmonics. The integration scheme, work flow, and application examples of the fault diagnosis system are presented. Simulation analysis and results show that the developed method can achieve effective diagnosis of the fault and fault degree of axle box bearings as well as yield better correction accuracy than the commonly used discrete spectrum correction methods.